AI Tools Could Catch Diabetes Years Before Blood Tests

Imagine finding out you're on track for diabetes not from a blood test, but from your smartwatch or a simple heart scan at your annual checkup. That future just got a lot closer.

Researchers at Stanford University and Imperial College London are developing AI-powered tools that can spot diabetes risk years before traditional blood tests show anything wrong. The breakthrough could help millions of people worldwide who are unknowingly living with the condition or headed toward it.

The numbers tell a sobering story. In the US alone, 11 million people have diabetes without knowing it. Another 115 million have prediabetes, and 80 percent are completely unaware. In the UK, up to 1.3 million people remain undiagnosed.

The problem runs deeper than numbers. By the time traditional blood sugar tests catch diabetes, damage may already be accumulating silently in the heart, kidneys, eyes, and nerves. Earlier detection means a real chance to prevent those complications or avoid diabetes entirely.

Stanford's Michael Snyder leads a team using continuous glucose monitors, the wearable sensors that track blood sugar in real time. Their AI algorithm analyzes patterns in the data to identify different forms of Type 2 diabetes with 90 percent accuracy.

"The goal from our standpoint is to keep people healthy versus try to fix them later," says Snyder, who developed Type 2 diabetes himself despite not fitting the typical profile. The monitors are becoming cheaper and more accessible, with many now available over the counter.

Meanwhile, researchers at Imperial College London took a completely different approach. They trained an AI system called AIRE-DM to read ordinary electrocardiograms, the simple heart tracings doctors already use daily.

Using 1.2 million ECGs from hospital records, the tool learned to spot subtle cardiovascular changes linked to future diabetes risk. It predicts who will develop the disease years later with 70 percent accuracy, matching or beating current diagnostic tools.

The beauty of this approach is scale. ECGs happen millions of times daily in hospitals and clinics worldwide. If approved for clinical use, AIRE-DM could automatically flag at-risk patients during routine care, no extra tests needed.

The Ripple Effect

These breakthroughs could reshape how we think about diabetes prevention. Instead of waiting for blood sugar to cross a threshold, doctors could intervene years earlier with lifestyle changes or medication. That head start matters enormously for the heart, kidneys, and other organs vulnerable to long-term damage.

The tools also promise to address troubling gaps in current testing. Recent studies show standard blood tests can read falsely low in some Black and South Asian people, delaying their diagnoses until the disease is more advanced. AI systems trained on diverse populations could help close that gap.

As these technologies become cheaper and more accessible, they could shift diabetes care from reactive to preventive, catching problems when they're easiest to stop.

The researchers envision a future where everyone wears a glucose monitor once a year or gets flagged during a routine heart check, turning silent progression into early warning and real prevention.